CN111897298A - Method and system for monitoring acoustic emission in preparation process of traditional Chinese medicine particles in fluidized bed - Google Patents

Abstract

The invention discloses an acoustic emission monitoring system for preparing traditional Chinese medicine particles by a fluidized bed, which comprises an acoustic emission sensor arranged on the bed body of the fluidized bed and used for receiving acoustic signal data of the fluidized bed; the amplifier and the acquisition card are used for amplifying and acquiring acoustic signal data at fixed time intervals; a computer system comprising a data preprocessing program module and a multivariate statistical process monitoring program module.

Description

Acoustic emission monitoring method and system for preparation process of traditional Chinese medicine granules in a fluidized bed

technical field

The invention relates to the technical field of on-line monitoring of the production process of traditional Chinese medicine granules, in particular to a process monitoring method and system for preparing traditional Chinese medicine granules in a fluidized bed based on acoustic emission technology.

Background technique

Fluidized bed spray granulation technology is a new type of granulation technology that integrates mixing, granulation and drying operations in one step. Compared with other methods, it has the advantages of simple process, short operation time and low labor intensity. In the pharmaceutical industry, the fluidized bed one-step granulation technology has been widely used at home and abroad.

The quality of traditional Chinese medicine granules has always been the focus of attention. The quality of the preparation process is an important factor affecting the quality of traditional Chinese medicine granules. The key to ensuring the stability of the granule production process is to monitor and control the granulation process in real time, so as to ensure a consistent production process and ultimately obtain high-quality products. The traditional process control strategy is to take samples at intervals during the granulation process and adjust the process parameters by observing or analyzing the samples offline. The offline detection method is cumbersome to operate, and the detection is hysteretic, which cannot meet the requirements of industrial production automation. Process monitoring technology can perform real-time online measurement of the fluidized bed granulation process and visualize the pharmaceutical granulation process, which is of great significance for deepening the understanding of the pharmaceutical process, controlling the quality of drugs and ensuring the consistency of drugs.

SUMMARY OF THE INVENTION

In view of the above-mentioned defects of the prior art, the technical problem to be solved by the present invention is how to monitor the preparation process of the fluidized bed Chinese medicine granules in a real-time, safe and environment-friendly manner.

In order to achieve the above object, the present invention provides, in the first aspect, a method for monitoring acoustic emission during the preparation of fluidized bed Chinese medicine granules, comprising the steps of:

(1) Acoustic signal data is collected by an acoustic emission sensor arranged on the fluidized bed;

(2) Take the collected acoustic signal data of the fluidized bed under normal working conditions as a calibration set; convert the acoustic signal data of the calibration set from time domain signals to frequency domain signals through fast Fourier transform, and convert the spectral averages. It is divided into N segments, and the average value of each segment is calculated to obtain a segmented average spectrum containing N variables; the data alignment algorithm is used to align the spectrum data;

(3) Establish a multivariate statistical process monitoring model including principal component control model, Hotelling's T2 control model and DModX control model; Ingredient control model, Hotelling's T2 control model and DModX control model for training;

(4) The multivariate statistical process monitoring model obtained by training is used to monitor the preparation process of TCM granules in the fluidized bed.

Further, in step (1), the acoustic emission sensor is set at the height of the sampling port of the charging chamber of the fluidized bed.

Further, in step (2), a correlation coefficient correction method is used to align the spectral data.

Further, in step (3), the calculation formula of the principal component score trajectory of the principal component control model is:

为NOC批次的载荷矩阵，E_NOC为残差矩阵；where T _NOC is the score matrix of NOC batches,

is the loading matrix of the NOC batch, and E _NOC is the residual matrix;

Control limits for the principal components control chart of the principal component control model using the mean ± 3 standard deviations.

Further, in step (3), the calculation formula of Hotelling's T ² control model is:

为t_i的方差，A表示主成分控制模型中主成分的数量；Hotelling's T²的控制限用用F分布来计算，公式为：where t _i is the score of the i-th principal component,

is the variance of t _i , A represents the number of principal components in the principal component control model; the control limit of Hotelling's T ² is calculated by F distribution, and the formula is:

In the formula, K is the number of batches of the calibration set, F is the critical value of the F distribution when the confidence level is 1-α and the degree of freedom is (A, K-A).

Further, in step (3), the DModX value calculation formula of any batch of acoustic signal data in the DModX control model at time k is:

为主成分控制模型中n时刻x_n的预测值，K为光谱变量，A表示主成分控制模型中主成分的数量；In the formula, e _n is the residual vector of the observation value x _n ,

is the predicted value of x _n at time n in the principal component control model, K is the spectral variable, and A represents the number of principal components in the principal component control model;

The control limits of the DModX control model were established by the mean ± 3 times the standard deviation, and the formula was:

为校正集的平均值，SD为建立主成分控制模型中样本DModX统计量的标准偏差。In the formula,

is the mean of the calibration set, SD is the standard deviation of the sample DModX statistic in the establishment of the principal component control model.

The present invention provides, in a second aspect, an acoustic emission monitoring system for preparing traditional Chinese medicine particles in a fluidized bed, comprising an acoustic emission sensor disposed on the fluidized bed body for receiving acoustic signal data of the fluidized bed; an amplifier and a collection card , used to amplify and collect acoustic signal data at fixed time intervals; the computer system includes a data preprocessing program module and a multivariate statistical process monitoring program module, wherein the data preprocessing program module converts the acoustic signal data from the The time domain signal is converted into a frequency domain signal, and the spectrum is divided into N segments, and the average value of each segment is calculated to obtain a segmented average spectrum containing N variables; the data alignment algorithm is used to align the spectrum data; multivariate statistical process The monitoring program modules include principal component control module, Hotelling's T ² control module and DModX control module.

Further, the calculation formula of the principal component score trajectory in the principal component control module is:

为NOC批次的载荷矩阵，E_NOC为残差矩阵；where T _NOC is the score matrix of NOC batches,

is the loading matrix of the NOC batch, and E _NOC is the residual matrix;

Control limits for the principal components control chart of the principal component control model using the mean ± 3 standard deviations.

Further, the calculation formula of Hotelling's T ² control module is:

为t_i的方差，A表示主成分控制模型中主成分的数量；Hotelling's T²的控制限用用F分布来计算，公式为：where t _i is the score of the i-th principal component,

is the variance of t _i , A represents the number of principal components in the principal component control model; the control limit of Hotelling's T ² is calculated by F distribution, and the formula is:

In the formula, K is the number of batches of the calibration set, F is the critical value of the F distribution when the confidence level is 1-α and the degree of freedom is (A, K-A).

Further, the calculation formula of the DModX value at time k of any batch of acoustic signal data in the DModX control module is:

为主成分控制模型中n时刻x_n的预测值，K为光谱变量，A表示主成分控制模型中主成分的数量；In the formula, e _n is the residual vector of the observation value x _n ,

is the predicted value of x _n at time n in the principal component control model, K is the spectral variable, and A represents the number of principal components in the principal component control model;

The control limits of the DModX control model were established by the mean ± 3 times the standard deviation, and the formula was:

校正集的平均值，SD为建立主成分控制模型中样本DModX统计量的标准偏差。In the formula,

The mean of the calibration set, SD is the standard deviation of the sample DModX statistic in the establishment of the principal component control model.

Acoustic emission detection technology is safe and environmentally friendly, sensitive in detection, rich in information, and the non-invasive signal acquisition method can ensure that the fluidized state in the process of fluidized bed granulation is not destroyed, and the acquisition speed of acoustic emission signals is fast. The process of data processing and statistical analysis can be completed in less than 1 minute. The use of acoustic emission to monitor the fluidized bed granulation process online can meet the requirements of real-time online measurement in industrial production, which is conducive to timely detection of product quality changes and timely regulation. , to maintain the stability of product quality.

The concept, specific structure and technical effects of the present invention will be further described below in conjunction with the accompanying drawings, so as to fully understand the purpose, characteristics and effects of the present invention.

Description of drawings

Fig. 1 is the flow chart of the acoustic emission monitoring method of the fluidized bed Chinese medicine particle preparation process in a preferred embodiment of the present invention;

Fig. 2 is the schematic diagram of the acoustic emission monitoring system of the preparation process of fluidized bed Chinese medicine granules in a preferred embodiment of the present invention;

3 is a spectrogram of different acquisition time lengths of acoustic signals in a preferred embodiment of the present invention;

Fig. 4 is the original average spectrogram of the acoustic signal in a preferred embodiment of the present invention;

Fig. 5 is the piecewise average frequency spectrum of the acoustic signal in a preferred embodiment of the present invention;

6 is a process data diagram before alignment processing in a preferred embodiment of the present invention;

7 is a process data diagram after alignment processing in a preferred embodiment of the present invention.

Detailed ways

The following describes several preferred embodiments of the present invention with reference to the accompanying drawings, so as to make its technical content clearer and easier to understand. The present invention can be embodied in many different forms of embodiments, and the protection scope of the present invention is not limited to the embodiments mentioned herein.

As shown in Figures 1 and 2, in another specific embodiment according to the present invention, 1 is a filter bag of the fluidized bed, 2 is a spray gun, 3 is a sampling port, 4 is an acoustic emission sensor, 5 is an amplifier, and 6 is an The acquisition card, 7 is the computer system. The acoustic emission sensor 4 is arranged at the height of the sampling port 3 of the fluidized bed charging chamber, and the vertical distance from the gas distribution plate is 23.5 cm. The acoustic signal is received by the sensor 4 and converted into an electrical signal, amplified by the amplifier 5 and converted into a digital signal by the DS5-8B full-information acoustic emission signal analyzer, and finally the acoustic signal data is recorded through the matching acquisition card 6 and acquisition software. The sampling frequency of the acoustic emission signal is 3MHz, and the acoustic signal is collected at fixed time intervals during the fluidized bed granulation process until the end of the granulation.

Acoustic emission signals during the preparation of traditional Chinese medicine granules in different batches of fluidized bed were collected, and the collected data of different batches were divided into calibration set and validation set according to a certain proportion.

In view of the fact that the pre-experiment shows that the spectral peak shapes of different acquisition durations are similar, the signal acquisition duration is set to 2s, as shown in Figure 3.

Acoustic emission signals during the fluidized bed granulation process were collected until the end of granulation. Before modeling, the original time-domain signal collected is converted into a frequency-domain signal. The specific scheme is to divide the 2s time-domain signal into 50 segments, and each segment of the time-domain signal is subjected to a fast Fourier transform (FFT) to obtain a spectrum. Average the 50 spectrums to obtain the average spectrogram of the 2s acoustic signal, as shown in Figure 4. The frequency range of the spectrum is 50-400kHz, the resolution is 25Hz, and each spectrum contains 14001 frequency variables. In this study, a segmented average spectrum with 700 variables was selected to build the model, as shown in Figure 5. Acoustic signal processing was done in MATLAB 2019b (MathWorks, USA).

In view of the problems of uneven sampling time interval and deviation of data collection time points in the acoustic emission signals obtained in the process of fluidized bed granulation, the correlation coefficient correction method is used to align the process data, and the spectrum before and after preprocessing is used to align the process data. Figures are shown in Figures 6 and 7. In the process of correlation coefficient correction, the endpoints of the spectral vectors to be aligned are fixed, and the spectral vectors are divided into the same number of segments as the reference spectral vectors according to the relaxation parameters. The last segment is compared and corrected with the reference spectral vectors. Adjust the data segments to be aligned by stretching or compressing transformation in the backward range, so as to obtain a set of data vectors with the largest correlation coefficient, and so on, and finally obtain a set of aligned recombination signal vectors, the calculation formula of the correlation coefficient as follows:

(I×J×K，I为实验批次，J为信号强度，K为时间)，按变量方向进行展开，形成I×K行、J列的二维矩阵X(IK×J)；Acoustic emission 3D data obtained during fluidized bed granulation

(I×J×K, I is the experimental batch, J is the signal intensity, and K is the time), expand according to the variable direction to form a two-dimensional matrix X (IK×J) with I×K rows and J columns;

A multivariate statistical process monitoring model was established using the acoustic emission signals of the fluidized bed granulation process of Yangwei granules under six normal operating conditions.

Multivariate statistical process monitoring models include principal components control charts, Hotelling's T2 control charts, and ^DModX control charts.

Among them, the principal component score trajectory can intuitively characterize the changes in the process, which can be used for monitoring and prediction of the granulation process. The calculation formula of the principal component score trajectory is:

为NOC批次的载荷矩阵，E_NOC为残差矩阵；采用均值±3倍标准偏差为主成分控制模型的主成分控制图的控制限。where T _NOC is the score matrix of NOC batches,

is the loading matrix of the NOC batch, and E _NOC is the residual matrix; the control limits of the principal component control chart of the principal component control model using the mean ± 3 times the standard deviation.

In Hotelling's T ² statistical process monitoring model, Hotelling's T ² statistic is the Mahalanobis distance from the sample to the origin in the principal component space, which can reflect the change of the variable through the fluctuation of the principal component vector inside the principal component model. The calculation formula is: :

为t_i的方差，A表示主成分控制模型中主成分的数量；Hotelling's T²的控制限用用F分布来计算，公式为：where t _i is the score of the i-th principal component,

is the variance of t _i , A represents the number of principal components in the principal component control model; the control limit of Hotelling's T ² is calculated by F distribution, and the formula is:

In the formula, K is the number of batches in the calibration set, F is the critical value of the F distribution when the confidence level is 1-α and the degree of freedom is (A, K-A).

In the DModX statistical process monitoring model, the DModX statistic is the residual standard deviation, that is, the absolute distance from the observed value to the model, which reflects the degree of change in the external data of the model. The calculation formula of the DModX value of any batch at time k is:

为所述主成分控制模型中n时刻x_n的预测值，K为光谱变量，A表示所述的主成分控制模型中主成分的数量；In the formula, e _n is the residual vector of the observation value x _n ,

is the predicted value of x _n at time n in the principal component control model, K is a spectral variable, and A represents the number of principal components in the principal component control model;

为所述校正集的平均值，SD为建立所述的主成分控制模型中样本DModX统计量的标准偏差。in the formula

is the mean value of the calibration set, SD is the standard deviation of the sample DModX statistic in establishing the principal component control model.

Substitute the calibration set data into Equation (2), Equation (3), and Equation (5) to calculate the corresponding monitoring indicators, and establish the principal component control model, Hotelling's T ² control model and DModX control model.

Substitute the monitoring data of the verification set and the process operating state into equations (4) and (6) to calculate the corresponding statistics, and monitor the fluidized granulation state of the new batch. For a new inspected batch, if the process trajectory falls within the control limits of the established multivariate statistical process control chart, the batch is considered to be in a normal state, otherwise, the batch is considered to be in an abnormal state.

The preferred embodiments of the present invention have been described in detail above. It should be understood that many modifications and changes can be made according to the concept of the present invention by those skilled in the art without creative efforts. Therefore, all technical solutions that can be obtained by those skilled in the art through logical analysis, reasoning or limited experiments on the basis of the prior art according to the concept of the present invention shall fall within the protection scope determined by the claims.

Claims (10)

1. a fluidized bed Chinese medicine particle preparation process acoustic emission monitoring method, is characterized in that, comprises the steps: (1) Acoustic signal data is collected by an acoustic emission sensor arranged on the fluidized bed; (2) Use the collected acoustic signal data of the fluidized bed in a normal working state as a calibration set; convert the acoustic signal data of the calibration set from a time domain signal to a frequency domain through fast Fourier transform The spectrum is divided into N segments, the average value of each segment is calculated, and the segmented average spectrum containing N variables is obtained; the spectrum data is aligned by the data alignment algorithm; (3) establishing a multivariate statistical process monitoring model including a principal component control model, a Hotelling's T2 control model and a ^DModX control model; by including at least 6 batches of all the acoustic signal data of the fluidized bed under normal working conditions The calibration set is used to train the principal component control model, Hotelling's T2 control model and ^DModX control model; (4) The multivariate statistical process monitoring model obtained by training is used to monitor the preparation process of TCM granules in the fluidized bed. 2. The method for monitoring acoustic emission in the preparation process of traditional Chinese medicine particles in a fluidized bed as claimed in claim 1, wherein, in step (1), the acoustic emission sensor is set at the sampling port of the charging chamber of the fluidized bed height. 3 . The method for monitoring acoustic emission in the preparation process of fluidized bed Chinese medicine particles according to claim 2 , wherein, in step (2), a correlation coefficient correction method is used to align the spectral data. 4 . 4. The method for monitoring acoustic emission in the preparation process of fluidized bed Chinese medicine granules as claimed in claim 3, wherein, in step (3), the calculation formula of the principal component score trajectory of the principal component control model is:

where T _NOC is the score matrix of NOC batches,

is the loading matrix of the NOC batch, and E _NOC is the residual matrix; The mean ± 3 times the standard deviation was used as the control limit of the principal component control chart of the principal component control model. 5. The method for monitoring acoustic emission in the preparation process of fluidized bed Chinese medicine particles as claimed in claim 4, wherein, in step (3), the calculation formula of described Hotelling's T ² control model is:

where t _i is the score of the i-th principal component,

is the variance of t _i , A represents the number of principal components in the described principal component control model; the control limit of Hotelling's T ² is calculated with F distribution, and the formula is:

In the formula, K is the batch number of the calibration set, F is the critical value of the F distribution when the confidence level is 1-α and the degree of freedom is (A, K-A). 6. The method for monitoring acoustic emission of fluidized bed Chinese medicine particle preparation process as claimed in claim 4, wherein, in step (3), the DModX value of the acoustic signal data of any batch in the described DModX control model at time k The calculation formula is:

In the formula, e _n is the residual vector of the observation value x _n ,

is the predicted value of x _n at time n in the principal component control model, K is a spectral variable, and A represents the number of principal components in the principal component control model; The control limits of the DModX control model were established by the mean ± 3 times the standard deviation, and the formula was:

In the formula,

is the mean value of the calibration set, SD is the standard deviation of the sample DModX statistic in establishing the principal component control model. 7. an acoustic emission monitoring system for preparing traditional Chinese medicine particles in a fluidized bed, is characterized in that, comprising The acoustic emission sensor arranged on the fluidized bed body is used to receive the acoustic signal data of the fluidized bed; Amplifier and acquisition card for amplifying and acquiring the acoustic signal data at fixed time intervals; A computer system, the computer system includes a data preprocessing program module and a multivariate statistical process monitoring program module, wherein the data preprocessing program module converts the acoustic signal data from a time domain signal to a fast Fourier transform. frequency domain signal, and divide the spectrum into N segments, and calculate the average value of each segment to obtain a segmented average spectrum containing N variables; use the data alignment algorithm to align the spectrum data; The multivariable statistical process monitoring program module includes a principal component control module, a Hotelling's T2 control module and a ^DModX control module. 8. The acoustic emission monitoring system for preparing traditional Chinese medicine particles in a fluidized bed as claimed in claim 7, wherein the calculation formula of the principal component score trajectory in the principal component control module is:

where T _NOC is the score matrix of NOC batches,

is the loading matrix of the NOC batch, and E _NOC is the residual matrix; The mean ± 3 times the standard deviation was used as the control limit of the principal component control chart of the principal component control model. 9. the acoustic emission monitoring system of fluidized bed preparation Chinese medicine granules as claimed in claim 7, wherein, the calculation formula of described Hotelling's T ² control module is:

where t _i is the score of the i-th principal component,

is the variance of t _i , A represents the number of principal components in the described principal component control model; the control limit of Hotelling's T ² is calculated with F distribution, and the formula is:

In the formula, K is the batch number of the calibration set, F is the critical value of the F distribution when the confidence level is 1-α and the degree of freedom is (A, K-A). 10. The acoustic emission monitoring system of fluidized bed preparation of traditional Chinese medicine particles as claimed in claim 7, wherein, the DModX value calculation formula of the acoustic signal data of any batch in the described DModX control module at time k is:

In the formula, e _n is the residual vector of the observation value x _n ,

is the predicted value of x _n at time n in the principal component control model, K is a spectral variable, and A represents the number of principal components in the principal component control model; The control limits of the DModX control model were established by the mean ± 3 times the standard deviation, and the formula was:

In the formula,

is the mean value of the calibration set, SD is the standard deviation of the sample DModX statistic in establishing the principal component control model.

Images